What happened to Helicone?

On 2026-10-31 the banner on helicone.ai read "Helicone is no longer part of Mintlify. Details to come soon.". The Reddit thread that ranks for this query is titled "Helicone was acquired by Mintlify"; what follows the banner is unannounced on Helicone's pages as this page is written. Its product pages, pricing and docs were all still up on the read.

Is there a free Helicone alternative?

Helicone's own Hobby plan is free (10,000 requests, 7-day retention, read 2026-10-31). Here the observability is part of the key rather than a plan: sessions, the usage CSV and the log exports carry no fee, and what you pay is the tokens, prepaid from $10.

Is there an open-source Helicone alternative on GitHub?

Helicone itself documents Docker Compose self-hosting, and Langfuse and OpenLIT are open-source observability backends. This key exports to Langfuse and to any OTLP/HTTP backend, so a self-hosted Langfuse can be the store while the gateway stays hosted.

How does this compare with Portkey or Langfuse?

Portkey is a gateway over provider keys you hold, priced by logged requests; Langfuse is an observability backend with no gateway. This is a hosted gateway with its own balance and models, whose logs can be sent to Langfuse. The OpenRouter alternatives page puts Portkey and the other gateways on $100 of usage.

Does the key keep my prompts?

No, unless you ask it to. Request and response bodies are not stored; a key created with retain_bodies on keeps them for 7 days so an export with include_bodies can carry them. Everything else exported is metering: model, tokens, cost, latency, outcome.

What does a session cost here?

The tokens only. Forty turns of 9,000 tokens in and 700 out on Claude Sonnet 5.5 come to $0.516 at today's catalog price ($1.20 in, $3.00 out per million tokens); the session timeline, the receipts and the exports add nothing.

Checked 2026-10-31

Helicone alternative: a gateway with the logs built in

Helicone sold observability in front of your own provider keys, then a gateway with credits. On 2026-10-31 its site carries a banner about its next chapter. RelaySwitch is the other shape: a hosted gateway with its own balance and models, where sessions, log export and a receipt per request come with the key.

A Helicone alternative is judged on what it logs, where it can send the logs, and whether the gateway in front is yours or theirs. Every Helicone fact below was read on 2026-10-31 from the page linked beside it; the RelaySwitch row is read from the live catalog and the OpenAPI document. The pillar for the family is the OpenRouter alternatives page.

Chat models here: 338
Under list, in and out: 70
Today's real-model check: 305 of 339 stable
First top-up: $10

01What Helicone is, and where it stands today

Helicone's home page (read 2026-10-31) is headed “Build Reliable AI Apps” and lists Requests, Sessions, Users, HQL, Prompts, Datasets, Playground, Monitor, Rate Limits and Alerts in its dashboard. Its AI Gateway is “a unified API for 100+ LLM providers through the OpenAI SDK format” at https://ai-gateway.helicone.ai, paid with credits at “0% markup” or with your own provider keys. Plans on its pricing page: Hobby free with 10,000 requests and 7 days of retention, Pro $79 a month with 1 month, Team $799 a month with 3 months, Enterprise on request with on-prem.

Across those pages on 2026-10-31 ran one banner: “Helicone is no longer part of Mintlify. Details to come soon.” What comes next is not announced on the site. Its strengths stand as read: routing syntax in the model string, a deep dashboard, prompts and datasets, and documented self-hosting.

02What you get here: gateway and observability on one key

Tag a run with x-session-id and every request of it lands on one timeline: model, tokens, cost, latency, outcome and tool calls, up to 1,000 requests a session.

python

1from openai import OpenAI

3client = OpenAI(base_url="https://api.relayswitch.io/v1", api_key="sk-relay-...")

4run_id = "triage-2026-10-31-0912"

5for step in plan:

6 reply = client.chat.completions.create(

7 model="claude-sonnet-5-5",

8 messages=history + [step],

9 extra_headers={"x-session-id": run_id},

10 )

Send the log where you already look. Exports are created once with an admin key and run on their own: a webhook (signed JSON batches), Langfuse (each request a generation, grouped by session) or any OTLP/HTTP backend (a session is one trace, each request a span). The usage log also downloads as CSV.

bash

1# A webhook destination: signed JSON batches of every request (admin key).

2curl https://api.relayswitch.io/v1/exports \

3 -H "Authorization: Bearer $ADMIN_KEY" -H "Content-Type: application/json" \

4 -d '{"kind": "webhook", "name": "prod log", "url": "https://example.com/llm-log"}'

6# The usage log as CSV, up to 92 days a file.

7curl "https://api.relayswitch.io/v1/usage/export?from=2026-10-01&to=2026-10-31" \

8 -H "Authorization: Bearer $ADMIN_KEY" -o usage.csv

Every request has a signed receipt with the model, the billed tokens and the charge; the Verify page checks the signature.

bash

1curl https://api.relayswitch.io/v1/requests/REQUEST_ID/receipt \

2 -H "Authorization: Bearer $RELAYSWITCH_API_KEY"

Session caps and a loop guard on the key stop a runaway agent before it spends the budget. The observability page has the fields, the export kinds and a session priced live.

03Side by side

	Helicone	RelaySwitch
Hosting	Hosted at helicone.ai; Docker Compose self-hosting documented; on-prem on Enterprisesource, read 2026-10-31	Hosted; the API, the console and the logs live on one account
Logs	Per plan: Hobby 10,000 free requests with 7-day retention, Pro $79 a month with 1 month, Team $799 a month with 3 months, Enterprise configurablesource, read 2026-10-31	One usage record per request, kept as an accounting record; session timelines up to 1,000 requests; bodies only on a key with retain_bodies on, for 7 days
Exports	Webhooks on completed requests (request_id, request and response bodies cut at 10 KB; model, provider, cost and latency with Include Data on); a data-export row on the pricing pagesource, read 2026-10-31	Signed webhook batches (HMAC-SHA256), Langfuse, any OTLP/HTTP backend (Datadog, Honeycomb, Grafana), and the usage log as CSV up to 92 days a file
Prepaid key	AI Gateway at ai-gateway.helicone.ai in the OpenAI SDK format: credits with provider keys managed for you at "0% markup", or BYOKsource, read 2026-10-31	Prepaid from $10, no subscription; most chat models with a list price sit under it; OpenAI and Anthropic formats on the same key
Crypto	No crypto top-up on its pricing pagesource, read 2026-10-31	USDT, USDC, BTC, ETH, SOL to the same USD balance, nothing added on the top-up, no identity check
Proof	No public model check on the pages readsource, read 2026-10-31	A daily canary per model, public at /proof, and a signed receipt per request

Helicone cells are what its linked pages state on 2026-10-31; where a page gives no figure the cell says so. RelaySwitch cells are from the live OpenAPI document and the catalog (70 of 84 chat models with a list price sit under it on input and output today).

The short version

Want a dashboard over provider keys you already hold and are happy to self-host: Helicone, as its docs stand today. Want the gateway, the balance, the models and the logs on one account, with exports to your own backend and a receipt per request: this key. Paid by Card, USDT, USDC, BTC, ETH, SOL, no identity check.

Questions

What happened to Helicone?: On 2026-10-31 the banner on helicone.ai read "Helicone is no longer part of Mintlify. Details to come soon.". The Reddit thread that ranks for this query is titled "Helicone was acquired by Mintlify"; what follows the banner is unannounced on Helicone's pages as this page is written. Its product pages, pricing and docs were all still up on the read.
Is there a free Helicone alternative?: Helicone's own Hobby plan is free (10,000 requests, 7-day retention, read 2026-10-31). Here the observability is part of the key rather than a plan: sessions, the usage CSV and the log exports carry no fee, and what you pay is the tokens, prepaid from $10.
Is there an open-source Helicone alternative on GitHub?: Helicone itself documents Docker Compose self-hosting, and Langfuse and OpenLIT are open-source observability backends. This key exports to Langfuse and to any OTLP/HTTP backend, so a self-hosted Langfuse can be the store while the gateway stays hosted.
How does this compare with Portkey or Langfuse?: Portkey is a gateway over provider keys you hold, priced by logged requests; Langfuse is an observability backend with no gateway. This is a hosted gateway with its own balance and models, whose logs can be sent to Langfuse. The OpenRouter alternatives page puts Portkey and the other gateways on $100 of usage.
Does the key keep my prompts?: No, unless you ask it to. Request and response bodies are not stored; a key created with retain_bodies on keeps them for 7 days so an export with include_bodies can carry them. Everything else exported is metering: model, tokens, cost, latency, outcome.
What does a session cost here?: The tokens only. Forty turns of 9,000 tokens in and 700 out on Claude Sonnet 5.5 come to $0.516 at today's catalog price ($1.20 in, $3.00 out per million tokens); the session timeline, the receipts and the exports add nothing.

Gateway and logs on one key

Prepaid from $10 by card or crypto. Sessions, exports and receipts included.

Observability Get a key